Novel quality evaluation method and system for first set of equipment in display industry

Abstract

The invention discloses a novel display industry first set of equipment quality evaluation method and system, and the method comprises the steps: obtaining the performance data and fault data of the first set of equipment of the novel display industry, and obtaining the historical performance data, historical fault data and historical production benefits of similar advanced equipment; comparing the performance data with the historical performance data to obtain change data, and performing feature selection on the change data to obtain change features; obtaining quality data according to the fault data, and obtaining prediction benefits based on the change characteristics, the quality data and historical production benefits; and constructing a comprehensive quality coefficient based on the change data, the quality data and the prediction benefit, establishing a quality evaluation model of the first set of equipment by adopting the comprehensive quality coefficient, and inputting the to-be-evaluated data into the quality evaluation model of the first set of equipment to output an evaluation result. The invention provides a scientific and effective novel display industry first set of equipment quality evaluation method, which is helpful for an enterprise to accurately evaluate the equipment quality, make a reasonable decision and promote industrial development.

Description

Technical Field

[0001] The present invention relates to the technical field of quality evaluation of new display industry equipment, and in particular to a quality evaluation method and system for the first set of equipment in the new display industry. Background Art

[0002] In the new display industry, the first set of equipment is a key driver of technological innovation and industrial upgrading. It usually integrates the latest technology and processes and is of great significance to the development of the industry.

[0003] However, the quality evaluation of the first set of equipment faces many challenges. Traditional quality evaluation methods are often based on existing standards and experience, which may not be applicable to the first set of new equipment. These methods may not accurately evaluate key indicators such as equipment performance, reliability and stability.

[0004] In addition, the innovation and uniqueness of the first set of equipment also makes its quality evaluation more complicated. Without an in-depth understanding of its special design and functions, as well as comparative analysis with similar advanced equipment, it is difficult to fully evaluate its quality level. At the same time, there is also a lack of effective prediction methods for the impact of the first set of equipment on production efficiency, which brings difficulties to the investment decision of enterprises.

[0005] Therefore, a new quality evaluation method is needed that can fully consider the characteristics of the first set of equipment, comprehensively evaluate its performance, reliability, production efficiency and other aspects, and provide enterprises with accurate decision-making basis. Summary of the invention

[0006] The purpose of the present invention is to provide a method and system for evaluating the quality of the first set of equipment in a new display industry.

[0007] To achieve the above object, the present invention is implemented according to the following technical solutions:

[0008] The present invention comprises the following steps:

[0009] Obtain performance data and failure data of the first set of equipment in the new display industry, and obtain historical performance data, historical failure data and historical production benefits of similar advanced equipment;

[0010] Comparing the performance data with the historical performance data to obtain change data, and performing feature selection on the change data to obtain change features;

[0011] Obtaining quality data according to the fault data, and obtaining predicted benefits based on the change characteristics, the quality data and the historical production benefits;

[0012] A comprehensive quality coefficient is constructed based on the change data, the quality data and the predicted benefit, a first-set equipment quality evaluation model is established using the comprehensive quality coefficient, and the data to be evaluated is input into the first-set equipment quality evaluation model to output an evaluation result.

[0013] Furthermore, the method for obtaining performance data and failure data of the first set of equipment in the new display industry includes:

[0014] Performance data includes production efficiency, parameters and performance of produced products, and failure data includes number of failures, failure time, failure repair cost and failure repair time.

[0015] Furthermore, the method of performing feature selection on the change data to obtain the change feature includes:

[0016] The change data is the percentage change of the performance data relative to the latest historical performance data. The historical performance data is used as the independent variable and the historical production benefits are used as the dependent variable to train the regression model:

[0017] The expected function of the predicted value of the dependent variable is:

[0018]

[0019] in, is the predicted value of the dependent variable, E[y|x] is the expected function of the predicted value of the dependent variable, f(x,y) is the joint probability density function of x and y, x is the independent variable, y is the dependent variable,

[0020] Estimate the joint probability density function using nonparametric estimation:

[0021]

[0022] Where i is the index of the independent and dependent variables, n is the number of i, k is the dimension of x, ρ is the standard deviation of the independent variable x, T represents the transpose of the matrix, and w(x i ) is the weight function, λ and μ are weight adjustment parameters;

[0023] The output formula for the predicted value of the dependent variable is:

[0024]

[0025] Feature importance of feature data of changing data:

[0026]

[0027] in For the The feature importance of the performance data, In the regression model The SHAP value of the historical performance data corresponding to the performance data. For the The change data of performance data, For the Time series data of historical performance data corresponding to the performance data, is the time series data of historical production benefits, for and The covariance of and They are and The standard deviation of , q1, q2 and q3 are the weight coefficients of SHAP value, change data and correlation respectively,

[0028] Feature data whose feature importance is less than the preset importance threshold are removed, and feature data whose feature importance is greater than or equal to the preset importance threshold are taken as change features. The preset importance threshold is 0.5.

[0029] Furthermore, the method for obtaining quality data according to the fault data includes:

[0030] The fault repair costs of the fault data and historical fault data are sorted in ascending order to obtain a cost series, and the fault repair time of the fault data and historical fault data is sorted in ascending order to obtain a time series. The ranking of the fault data and historical fault data in the cost series and the ranking in the time series are used to establish two-dimensional data points, and the K-means algorithm is used to cluster the data points:

[0031] Initially, each data point is considered as a cluster, and the two clusters with the smallest inter-cluster distance are merged in each iteration until the number of clusters is 3.

[0032] The data points are spaced:

[0033]

[0034] Where x and y are different data points. and x s are the rankings of data point x in the cost series and time series, respectively. and s are the rankings of data point y in the cost series and time series, respectively. and w s are the weights of failure repair cost and failure repair time, respectively.

[0035] The inter-cluster spacing is:

[0036]

[0037] Where D S (I m ,I n ) is I m and I n The inter-cluster distance, d S (x,y) is the distance between data points x and y, I m and I n For different clusters, |I m |and|I n |For I m and I n The number of data points included,

[0038] The three clusters correspond to three fault levels: low-level fault, medium-level fault, and high-level fault. The cluster with the largest average Euclidean length of data points is a high-level fault, and the cluster with the smallest average Euclidean length of data points is a low-level fault.

[0039] Extract the failure rate and mean time between failures of the failure data and historical failure data. The quality data is calculated as follows:

[0040]

[0041] Where Z z For quality data, is the mean time between failures, g l is the failure rate, c d 、c z and c g are the weights of low-level faults, medium-level faults, and high-level faults, respectively. and are the number of low-level faults, medium-level faults, and high-level faults, respectively.

[0042] Furthermore, a method for obtaining predicted benefits based on the change characteristics, the quality data and the historical production benefits includes:

[0043] The change characteristics and quality data of similar advanced equipment are used as input data, the corresponding historical production benefits are used as target data, and the GRU algorithm is used to train the prediction model. The objective function is:

[0044]

[0045] Where Θ represents the parameter set of the forecasting model, n is the number of samples, T is the sequence length of the time series data of historical production benefits, and y i,t is the i-th historical production benefit, is the prediction model for y i,t The predicted data, α is the coefficient of the L1 regularization term, ξ is a constant term used to ensure numerical stability, and ξ=1×10-8 ,

[0046] Split the update gate weight matrix in GRU into 2 temporary update gate matrices and update the output of the gate:

[0047]

[0048] in, and are the outputs of the first temporary update gate and the second temporary update gate at time t, respectively, t is the output of the update gate at time t, W z1 and W z2 are the first temporary update gate matrix and the second temporary update gate matrix, ∈ z1 and ∈ z2 are the noise terms of the first temporary update gate and the second temporary update gate, E is the spectrum width matrix of the multifractal spectrum of the input data, σ(·) is the sigmoid function, ⊙ represents the multiplication of the corresponding elements of the matrix, and h t-1 is the hidden layer state at time t-1, x t is the input data at time t,

[0049] The reset gate and update gate are used to update the hidden layer state to realize the long-term information memory of the GRU algorithm, and the convolutional neural network is used to fuse the hidden layer states at each moment to predict the prediction data of the prediction model:

[0050]

[0051] in and are the output results of the hidden layer of the recurrent neural network state fusion strategy for the i-th input data at time t and time t-1, respectively, t is the hidden layer state at time t, cat(·) represents the concatenation operation on the feature dimension, conv(·) represents the convolution operation with a convolution kernel size of 1,

[0052] Input the changing characteristics and quality data of the first set of equipment into the prediction model and output the predicted benefits.

[0053] Furthermore, a method for constructing a comprehensive quality coefficient based on the change data, the quality data and the predicted benefit includes:

[0054] Normalize the change data, quality data and predicted benefits, and calculate the comprehensive quality coefficient:

[0055]

[0056] Among them, Z s is the comprehensive quality coefficient of the first set of equipment, B μis the μth change data, is the weight of the μth change data, Z z The quality data of the first set of equipment, and is the weight of quality data and prediction benefit, is the predicted benefit of the first set of equipment at time t, is the sequence length for predicting benefits.

[0057] Furthermore, a method for establishing a first-set equipment quality evaluation model using the comprehensive quality coefficient includes:

[0058] A data set is established using change data, change characteristics, quality data and comprehensive quality coefficient. The random forest algorithm is used to take change data, change characteristics and quality data as input features and the comprehensive quality coefficient as output target to train the first set of equipment quality evaluation model.

[0059] A new type of display industry first set of equipment quality evaluation system, characterized by including:

[0060] The acquisition module obtains the performance data and fault data of the first set of equipment in the new display industry, and obtains the historical performance data, historical fault data and historical production benefits of similar advanced equipment;

[0061] a calculation module, comparing the performance data with the historical performance data to obtain change data, obtaining quality data according to the fault data, obtaining predicted benefits based on the change characteristics, the quality data and the historical production benefits, constructing a comprehensive quality coefficient based on the change data, the quality data and the predicted benefits, and establishing a quality evaluation model for the first set of equipment using the comprehensive quality coefficient;

[0062] A feature selection module performs feature selection on the change data to obtain change features;

[0063] The output module inputs the data to be evaluated into the first set of equipment quality evaluation model and outputs the evaluation results.

[0064] The beneficial effects of the present invention are:

[0065] The present invention comprehensively evaluates the quality of the first set of equipment by comprehensively considering multiple data; improves accuracy by focusing on key factors through feature selection; the established first set of equipment quality evaluation model can accurately output results, providing an effective tool for enterprise decision-making. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] Figure 1 This is a flow chart of the steps of a new display industry first set of equipment quality evaluation method and system of the present invention. DETAILED DESCRIPTION

[0067] The present invention is further described below by means of specific embodiments. The illustrative embodiments and descriptions of the present invention are used to explain the present invention but are not intended to limit the present invention.

[0068] The present invention provides a new type of display industry first set of equipment quality evaluation method and system thereof, comprising the following steps:

[0069] like Figure 1 As shown, in this embodiment, the following steps are included:

[0070] Obtain performance data and failure data of the first set of equipment in the new display industry, and obtain historical performance data, historical failure data and historical production benefits of similar advanced equipment;

[0071] In the actual evaluation, the performance data of the first set of equipment in the new display industry was obtained through the sensors and production recording system installed on it. The performance data are as follows: the production efficiency is 30 display panels per hour, the product resolution reaches 3840×2160, and the contrast is 10000:1; the failure data is: in the first 6 months of operation, the number of failures is 5 times, the average time for each failure is 2 hours, the average cost of each failure is 500 yuan, and the average time for each failure is 3 hours; the three-year historical performance data of similar advanced equipment is sorted in chronological order as follows: production efficiency [25, 26, 28], product resolution [1920×1080, 1920×1080, 2560×1440], contrast [7000:1, 8000:1, 9000:1]; historical failure data: the average number of failures is 10 times per year, the average time for each failure is 1.5 hours, the average maintenance cost is 400 yuan per time, and the average maintenance time is 2.5 hours per time; historical production efficiency: in the off-season, the monthly profit is [30,000, 40,000, 50,000] yuan, and in the peak season, the monthly profit is [60,000, 70,000, 80,000] yuan;

[0072] Comparing the performance data with the historical performance data to obtain change data, and performing feature selection on the change data to obtain change features;

[0073] In the actual evaluation, the change data are: production efficiency 7%, product resolution 125%, contrast 11%; the change characteristics are production efficiency, product resolution and contrast;

[0074] Obtaining quality data according to the fault data, and obtaining predicted benefits based on the change characteristics, the quality data and the historical production benefits;

[0075] In the actual evaluation, the quality data is: 0.7 for the first set of equipment, 0.5 for similar advanced equipment; the predicted benefits are: 75,000 yuan per month in the off-season, and 100,000 yuan per month in the peak season;

[0076] Constructing a comprehensive quality coefficient based on the change data, the quality data and the predicted benefit, establishing a first-set equipment quality evaluation model using the comprehensive quality coefficient, inputting the data to be evaluated into the first-set equipment quality evaluation model to output an evaluation result;

[0077] In the actual evaluation, the comprehensive quality coefficient of the first set of equipment was 0.75.

[0078] In this embodiment, the method for obtaining performance data and fault data of the first set of equipment in the new display industry includes:

[0079] Performance data includes production efficiency, parameters and performance of produced products, and failure data includes number of failures, failure time, failure repair cost and failure repair time.

[0080] In this embodiment, the method of performing feature selection on the change data to obtain the change feature includes:

[0081] The change data is the percentage change of the performance data relative to the latest historical performance data. The historical performance data is used as the independent variable and the historical production benefits are used as the dependent variable to train the regression model:

[0082] The expected function of the predicted value of the dependent variable is:

[0083]

[0084] in, is the predicted value of the dependent variable, E[y|x] is the expected function of the predicted value of the dependent variable, f(x,y) is the joint probability density function of x and y, x is the independent variable, y is the dependent variable,

[0085] Estimate the joint probability density function using nonparametric estimation:

[0086]

[0087] Where i is the index of the independent and dependent variables, n is the number of i, k is the dimension of x, ρ is the standard deviation of the independent variable x, T represents the transpose of the matrix, and w(x i ) is the weight function, λ and μ are weight adjustment parameters;

[0088] The output formula for the predicted value of the dependent variable is:

[0089]

[0090] Feature importance of feature data of changing data:

[0091]

[0092] in For the The feature importance of the performance data, In the regression model The SHAP value of the historical performance data corresponding to the performance data. For the The change data of performance data, For the Time series data of historical performance data corresponding to the performance data, is the time series data of historical production benefits, for and The covariance of and They are and The standard deviation of , q1, q2 and q3 are the weight coefficients of SHAP value, change data and correlation respectively,

[0093] Feature data whose feature importance is less than the preset importance threshold are removed, and feature data whose feature importance is greater than or equal to the preset importance threshold are taken as change features. The preset importance threshold is 0.5.

[0094] In this embodiment, the method for obtaining quality data according to the fault data includes:

[0095] The fault repair costs of the fault data and historical fault data are sorted in ascending order to obtain a cost series, and the fault repair time of the fault data and historical fault data is sorted in ascending order to obtain a time series. The ranking of the fault data and historical fault data in the cost series and the ranking in the time series are used to establish two-dimensional data points, and the K-means algorithm is used to cluster the data points:

[0096] Initially, each data point is considered as a cluster, and the two clusters with the smallest inter-cluster distance are merged in each iteration until the number of clusters is 3.

[0097] The data points are spaced:

[0098]

[0099] Where x and y are different data points. and x s are the rankings of data point x in the cost series and time series, respectively. and sare the rankings of data point y in the cost series and time series, respectively. and w s are the weights of failure repair cost and failure repair time, respectively.

[0100] The inter-cluster spacing is:

[0101]

[0102] Where D S (I m ,I n ) is I m and I n The inter-cluster distance, d S (x,y) is the distance between data points x and y, I m and I n For different clusters, |I m |and|I n |For I m and I n The number of data points included,

[0103] The three clusters correspond to three fault levels: low-level fault, medium-level fault, and high-level fault. The cluster with the largest average Euclidean length of data points is a high-level fault, and the cluster with the smallest average Euclidean length of data points is a low-level fault.

[0104] Extract the failure rate and mean time between failures of the failure data and historical failure data. The quality data is calculated as follows:

[0105]

[0106] Where Z z For quality data, is the mean time between failures, g l is the failure rate, c d 、c z and c g are the weights of low-level faults, medium-level faults, and high-level faults, respectively. and are the number of low-level faults, medium-level faults, and high-level faults, respectively.

[0107] In this embodiment, the method for obtaining predicted benefits based on the change characteristics, the quality data and the historical production benefits includes:

[0108] The change characteristics and quality data of similar advanced equipment are used as input data, the corresponding historical production benefits are used as target data, and the GRU algorithm is used to train the prediction model. The objective function is:

[0109]

[0110] Where Θ represents the parameter set of the prediction model, including the weight matrix, bias term, noise term and convolution kernel parameters, n is the number of samples, T is the sequence length of the time series data of historical production benefits, and y i,t is the i-th historical production benefit, is the prediction model for y i,t The predicted data, α is the coefficient of the L1 regularization term, ξ is a constant term used to ensure numerical stability, and ξ=1×10 -8 ,

[0111] Split the update gate weight matrix in GRU into 2 temporary update gate matrices and update the output of the gate:

[0112]

[0113] in, and are the outputs of the first temporary update gate and the second temporary update gate at time t, respectively, t is the output of the update gate at time t, W z1 and W z2 are the first temporary update gate matrix and the second temporary update gate matrix, ∈ z1 and ∈ z2 are the noise terms of the first temporary update gate and the second temporary update gate, E is the spectrum width matrix of the multifractal spectrum of the input data, σ(·) is the sigmoid function, ⊙ represents the multiplication of the corresponding elements of the matrix, and h t-1 is the hidden layer state at time t-1, x t is the input data at time t,

[0114] The reset gate and update gate are used to update the hidden layer state to realize the long-term information memory of the GRU algorithm, and the convolutional neural network is used to fuse the hidden layer states at each moment to predict the prediction data of the prediction model:

[0115]

[0116] in and are the output results of the hidden layer of the recurrent neural network state fusion strategy for the i-th input data at time t and time t-1, respectively, t is the hidden layer state at time t, cat(·) represents the concatenation operation on the feature dimension, conv(·) represents the convolution operation with a convolution kernel size of 1,

[0117] Input the changing characteristics and quality data of the first set of equipment into the prediction model and output the predicted benefits.

[0118] In this embodiment, the method for constructing a comprehensive quality coefficient based on the change data, the quality data and the predicted benefit includes:

[0119] Normalize the change data, quality data and predicted benefits, and calculate the comprehensive quality coefficient:

[0120]

[0121] Among them, Z s is the comprehensive quality coefficient of the first set of equipment, B μ is the μth change data, is the weight of the μth change data, Z z The quality data of the first set of equipment, and is the weight of quality data and prediction benefit, is the predicted benefit of the first set of equipment at time t, is the sequence length for predicting benefits.

[0122] In this embodiment, the method of establishing the first set of equipment quality evaluation model using the comprehensive quality coefficient includes:

[0123] A data set is established using change data, change characteristics, quality data and comprehensive quality coefficient. The random forest algorithm is used to take change data, change characteristics and quality data as input features and the comprehensive quality coefficient as output target to train the first set of equipment quality evaluation model.

[0124] A new type of display industry first set of equipment quality evaluation system, characterized by including:

[0125] The acquisition module obtains the performance data and fault data of the first set of equipment in the new display industry, and obtains the historical performance data, historical fault data and historical production benefits of similar advanced equipment;

[0126] a calculation module, comparing the performance data with the historical performance data to obtain change data, obtaining quality data according to the fault data, obtaining predicted benefits based on the change characteristics, the quality data and the historical production benefits, constructing a comprehensive quality coefficient based on the change data, the quality data and the predicted benefits, and establishing a quality evaluation model for the first set of equipment using the comprehensive quality coefficient;

[0127] A feature selection module performs feature selection on the change data to obtain change features;

[0128] The output module inputs the data to be evaluated into the first set of equipment quality evaluation model and outputs the evaluation results.

[0129] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A quality evaluation method for the first set of equipment in the new display industry, characterized in that: The following steps are involved: Obtain performance data and failure data of the first set of equipment in the new display industry, and obtain historical performance data, historical failure data and historical production benefits of similar advanced equipment; Comparing the performance data with the historical performance data to obtain change data, and performing feature selection on the change data to obtain change features; Obtaining quality data according to the fault data, and obtaining predicted benefits based on the change characteristics, the quality data and the historical production benefits; A comprehensive quality coefficient is constructed based on the change data, the quality data and the predicted benefit, a first-set equipment quality evaluation model is established using the comprehensive quality coefficient, and the data to be evaluated is input into the first-set equipment quality evaluation model to output an evaluation result.

2. According to claim 1, a method for evaluating the quality of the first set of equipment in the new display industry is characterized in that: Methods for obtaining performance data and failure data of the first set of equipment in the new display industry include: Performance data includes production efficiency, parameters and performance of produced products, and failure data includes number of failures, failure time, failure repair cost and failure repair time.

3. According to claim 1, a method for evaluating the quality of the first set of equipment in the new display industry is characterized in that: The method of performing feature selection on the change data to obtain the change feature comprises: The change data is the percentage change of the performance data relative to the latest historical performance data. The historical performance data is used as the independent variable and the historical production benefits are used as the dependent variable to train the regression model: The expected function of the predicted value of the dependent variable is: in, is the predicted value of the dependent variable, E[y|x] is the expected function of the predicted value of the dependent variable, f(x,y is the joint probability density function of x and y, x is the independent variable, y is the dependent variable, Estimate the joint probability density function using nonparametric estimation: Where i is the index of the independent and dependent variables, n is the number of i, k is the dimension of x, ρ is the standard deviation of the independent variable x, T represents the transpose of the matrix, and w(x i ) is the weight function, λ and μ are weight adjustment parameters; The output formula for the predicted value of the dependent variable is: Feature importance of feature data of changing data: in For the The feature importance of the performance data, In the regression model The SHAP value of the historical performance data corresponding to the performance data. For the The change data of performance data, For the Time series data of historical performance data corresponding to the performance data, is the time series data of historical production benefits, for and The covariance of and They are and The standard deviation of , q1, q2 and q3 are the weight coefficients of SHAP value, change data and correlation respectively, Feature data whose feature importance is less than the preset importance threshold are removed, and feature data whose feature importance is greater than or equal to the preset importance threshold are taken as change features. The preset importance threshold is 0.

5.

4. According to claim 1, a method for evaluating the quality of the first set of equipment in the new display industry is characterized in that: The method for obtaining quality data according to the fault data comprises: The fault repair costs of the fault data and historical fault data are sorted in ascending order to obtain a cost series, and the fault repair time of the fault data and historical fault data is sorted in ascending order to obtain a time series. The ranking of the fault data and historical fault data in the cost series and the ranking in the time series are used to establish two-dimensional data points, and the K-means algorithm is used to cluster the data points: Initially, each data point is considered as a cluster, and the two clusters with the smallest inter-cluster distance are merged in each iteration until the number of clusters is 3. The data points are spaced: Where x and y are different data points. and x s are the rankings of data point x in the cost series and time series, respectively. and s are the rankings of data point y in the cost series and time series, respectively. and w s are the weights of failure repair cost and failure repair time, respectively. The inter-cluster spacing is: Where D S (I m ,I n ) is I m and I n The inter-cluster distance, d S (x,y) is the distance between data points x and y, I m and I n For different clusters, |I m |and|I n |For I m and I n The number of data points included, The three clusters correspond to three fault levels: low-level fault, medium-level fault, and high-level fault. The cluster with the largest average Euclidean length of data points is a high-level fault, and the cluster with the smallest average Euclidean length of data points is a low-level fault. Extract the failure rate and mean time between failures of the failure data and historical failure data. The quality data is calculated as follows: Where Z z For quality data, is the mean time between failures, g l is the failure rate, c d 、c z and c g are the weights of low-level faults, medium-level faults, and high-level faults, respectively. and are the number of low-level faults, medium-level faults, and high-level faults, respectively.

5. According to claim 1, a method for evaluating the quality of the first set of equipment in the new display industry is characterized in that: The method for obtaining predicted benefits based on the change characteristics, the quality data and the historical production benefits includes: The change characteristics and quality data of similar advanced equipment are used as input data, the corresponding historical production benefits are used as target data, and the GRU algorithm is used to train the prediction model. The objective function is: Where Θ represents the parameter set of the forecasting model, n is the number of samples, T is the sequence length of the time series data of historical production benefits, and y i,t is the i-th historical production benefit, is the prediction model for y i,t The predicted data, α is the coefficient of the L1 regularization term, ξ is a constant term used to ensure numerical stability, and ξ=1×10 -8 , Split the update gate weight matrix in GRU into 2 temporary update gate matrices and update the output of the gate: in, and are the outputs of the first temporary update gate and the second temporary update gate at time t, respectively, t is the output of the update gate at time t, W z1 and W z2 are the first temporary update gate matrix and the second temporary update gate matrix, ∈ z1 and ∈ z2 are the noise terms of the first temporary update gate and the second temporary update gate, E is the spectrum width matrix of the multifractal spectrum of the input data, σ(·) is the sigmoid function, ⊙ represents the multiplication of the corresponding elements of the matrix, and h t-1 is the hidden layer state at time t-1, x t is the input data at time t, The reset gate and update gate are used to update the hidden layer state to realize the long-term information memory of the GRU algorithm, and the convolutional neural network is used to fuse the hidden layer states at each moment to predict the prediction data of the prediction model: in and are the output results of the hidden layer of the recurrent neural network state fusion strategy for the i-th input data at time t and time t-1, respectively, t is the hidden layer state at time t, cat(·) represents the concatenation operation on the feature dimension, conv(·) represents the convolution operation with a convolution kernel size of 1, Input the changing characteristics and quality data of the first set of equipment into the prediction model and output the predicted benefits.

6. According to claim 1, a method for evaluating the quality of the first set of equipment in the new display industry is characterized in that: The method for constructing a comprehensive quality coefficient based on the change data, the quality data and the predicted benefit includes: Normalize the change data, quality data and predicted benefits, and calculate the comprehensive quality coefficient: Among them, Z s is the comprehensive quality coefficient of the first set of equipment, B μ is the μth change data, is the weight of the μth change data, Z z The quality data of the first set of equipment, and is the weight of quality data and prediction benefit, is the predicted benefit of the first set of equipment at time t, is the sequence length for predicting benefits.

7. According to claim 1, a method for evaluating the quality of the first set of equipment in the new display industry is characterized in that: The method of establishing a first-set equipment quality evaluation model using the comprehensive quality coefficient includes: A data set is established using change data, change characteristics, quality data and comprehensive quality coefficient. The random forest algorithm is used to take change data, change characteristics and quality data as input features and the comprehensive quality coefficient as output target to train the first set of equipment quality evaluation model.

8. A quality evaluation system for the first set of equipment in a new display industry, used to implement the quality evaluation method for the first set of equipment in a new display industry as described in any one of claims 1 to 8, characterized in that: include: The acquisition module obtains the performance data and fault data of the first set of equipment in the new display industry, and obtains the historical performance data, historical fault data and historical production benefits of similar advanced equipment; a calculation module, comparing the performance data with the historical performance data to obtain change data, obtaining quality data according to the fault data, obtaining predicted benefits based on the change characteristics, the quality data and the historical production benefits, constructing a comprehensive quality coefficient based on the change data, the quality data and the predicted benefits, and establishing a quality evaluation model for the first set of equipment using the comprehensive quality coefficient; A feature selection module performs feature selection on the change data to obtain change features; The output module inputs the data to be evaluated into the first set of equipment quality evaluation model and outputs the evaluation results.

Images